Caywork Platform
Author at Caywork
If you built workflows on Agent.ai, you already know the platform is gone. What you're looking for now isn't marketing; it's a straight, line-by-line answer to "Where do I go next, and what does it actually cost me?" This guide puts Caywork next to Agent.ai across the four things that matter most in a switch: pricing, the no-code builder, the marketplace, and what happens to your work if you were a creator, not just a user. No filler, no hedging, just the comparison you'd want if you were doing this research yourself at midnight with a deadline.
What Happened to Agent.ai
Before you rebuild anything, it helps to know exactly what ended and why, because it changes what you should expect from wherever you land next. This wasn't a sudden failure or a funding collapse; it was a deliberate wind-down tied to how the platform was built and who it belonged to from day one. Understanding the mechanics of the shutdown also explains why there's no clean export button waiting for you and why creators specifically are the ones left with the most work to do. The short version: a side project ran its course, got folded into a bigger platform, and the parts that didn't fit, namely, the open creator marketplace, didn't come along for the ride.
Agent.ai was built by HubSpot co-founder Dharmesh Shah as a "professional network of AI agents," growing to more than 2 million users and over 44,000 agents built on the platform before it closed on August 22, 2026. Caywork's own coverage of the closure lays out three reasons: the platform was run as a side project rather than a revenue-focused product, HubSpot absorbed the underlying technology for rapid integration into its own roadmap, and standalone agent marketplaces in general have been losing ground to agent features embedded directly inside tools people already use. That last point matters; it's the same pressure every independent marketplace, Caywork included, has to answer for.
The clearest evidence of where things landed: as of this writing, agent.ai no longer resolves to the old marketplace at all; the domain redirects straight to BuilderPack, a HubSpot Labs product that ships AI actions built for HubSpot's Agent Builder, which entered public beta in July 2026. In other words, the brand didn't get relaunched: the technology got absorbed and the URL got repointed. And per Caywork's closure post, HubSpot never announced a migration path for creator listings, meaning if you published agents on Agent.ai, nobody is moving that work for you.
Agent.ai vs Caywork at a Glance
Before going section by section, it's worth seeing the shape of the decision in one place, because the two platforms aren't just different in degree; they're built on different assumptions about how people want to pay and who gets to publish. Agent.ai was free-to-start and creator-open by design; Caywork is pay-as-you-go and creator-open by design, which sounds similar until you look at what each one actually charges for. The table below isn't exhaustive, but it covers the four axes that drive most switching decisions: what you pay, how you build, how big the catalog is, and what a creator actually keeps. Read it as a starting point, then use the sections below to go deeper on whichever row matters most to your decision.
| Agent.ai (closed Aug 22, 2026) | Caywork | |
|---|---|---|
| Pricing model | Free pay-per-task credits, later layered with platform-based plans | One-time credit packs, no subscription, credits never expire |
| Builder | Web-based agent builder, Agent Teams (Pro-only) | No-code visual builder with Loop, Condition, and Delay nodes |
| Marketplace | 44,000+ agents built, 2M+ users | 120+ listed agents across 6 categories, 5,000+ tasks run |
| Creator payout | No public revenue-share program | 75% of every sale goes to the creator |
| Status | Closed August 22, 2026, absorbed into HubSpot | Active |
Pricing: Credits vs. Agent.ai's Legacy Tiers
Pricing is usually where a migration decision actually gets made, so it's worth being precise instead of hand-wavy about it. Agent.ai's original pitch was simplicity (pay per task, keep what you don't use), but that changed as the platform matured, and it's exactly the kind of thing that leaves users confused mid-switch about what they were even paying for. Caywork took a different, narrower bet: no subscriptions at all, in either direction. The two sections below walk through what each model actually meant for a working user, not just the headline pitch.
How Caywork's Credit System Works
Caywork sells credits in one-time packs rather than a recurring subscription, and the pricing page is explicit that this is deliberate: "no monthly fees, no hidden costs." There are three packs: a Starter pack at $5 for 100 credits, a Middle pack (the most popular option) at $15 for 300 credits, and a High pack at $45 for 900 credits. Credits are spent per agent run, at a cost the agent's own listing sets (most of the marketing agents we checked run 2-8 credits per task), and, importantly, credits don't expire, so buying ahead doesn't create a use-it-or-lose-it clock. There's no free tier in the traditional sense, but every agent can be explored before you commit to a purchase.
What Agent.ai Users Are Losing
Agent.ai launched on a pay-per-task credit model (one credit per task, free to start), which is part of why it grew as fast as it did. Later in its life, the platform introduced platform-based pricing tiers layered on top of that original free-credits system, a change that its own blog had to explain in detail because it confused a meaningful share of the user base. Whatever plan you were on when the shutdown hit, none of it carries forward: there's no account credit transfer, no grandfathered rate, and no discount code waiting on the other side. Whatever budget line you had for Agent.ai simply needs to be reopened wherever you land next.
Builder & No-Code Experience
Pricing gets you in the door, but the builder is what determines whether you can actually rebuild your workflows once you're there, and this is usually the part people underestimate when they're focused on cost. Agent.ai's builder was approachable but relatively flat: you could chain agents, but multi-step logic wasn't a first-class feature until late in the platform's life. Caywork's builder, especially after its most recent major release, leans harder into actual workflow logic rather than single-shot agent runs. The two subsections below cover the visual builder itself and the control-flow features that matter most if your old workflows had more than one step.
Visual Builder Comparison
Both platforms are built around the same core promise (no code required to assemble an agent), but they diverge on depth. Agent.ai's builder was designed for chaining pre-built agents into simple sequences, which covered a lot of use cases but hit a ceiling once a workflow needed branching logic or retries. Caywork's builder is organized around the same drag-and-connect model, but with explicit workflow nodes rather than a straight linear chain, which is the difference that shows up the moment you try to rebuild something more complex than a single-step agent.
Multi-Agent/Teamworkflows (V22 Loop, Condition, Delay Nodes)
Agent.ai's answer to multi-agent coordination was Agent Teams: a genuinely popular feature, but one that was gated behind its Pro tier and didn't survive the shutdown any more than anything else did. Caywork's V22 release shipped three control-flow nodes that cover most of the same ground without a separate paid tier: a LoopNode that repeats a set of steps until a condition is met (useful for processing lists of records or retrying a failed step), a ConditionNode that routes execution down different paths based on a previous step's outcome, and a DelayNode that pauses a workflow for a set amount of time, handy for scheduling follow-ups. If your Agent.ai setup relied on Agent Teams for anything beyond a simple two-step handoff, these three nodes are where you'll rebuild that logic.
Marketplace: 120+ Agents vs. Agent.ai's 44,000+ Built
Raw numbers can be misleading here, so it's worth separating what's actually being counted before you draw a conclusion from either figure. Agent.ai's 44,000+ figure counts everything ever built on the platform over roughly a year and a half, including one-off experiments and agents nobody but their creator ever used. Caywork's 120+ figure counts what's currently live and discoverable in the marketplace today. They're not directly comparable numbers, but they answer different useful questions: one tells you how much total creative activity a platform generated, and the other tells you how much is realistically browsable right now.
Category Coverage (Marketing, Finance, Education, E-commerce, Dev)
Caywork's marketplace is organized into six categories: Development, Marketing, Lifestyle, E-commerce, Education and Tutoring, and Finance and Banking. The marketing category alone includes named agents like Blog Post Writer: Custom Voice, SEO & GEO, OmniChannel Social Post Architect, SEO Growth Engine: Strategic Keyword Architect, and Conversational Closer: Sales Narrative Pro, a set that overlaps meaningfully with Agent.ai's most popular content and prospecting agents. If your old workflow lived in sales, marketing, or content, there's a reasonable chance a direct or close equivalent already exists; the complete migration map goes agent-by-agent if you want the specifics.
Quality, Ratings & Sandboxing
An open marketplace lives or dies on whether people trust what they're installing, and this is a documented weak point for agent marketplaces generally; supply-chain and trust concerns have been raised about third-party agent ecosystems more than once. Caywork's listings carry ratings and reviews as a baseline trust signal, the same community-feedback mechanism that helped Agent.ai's own top agents rise to the top of its catalog. If you're evaluating a third-party agent before connecting it to real business data, treat that the same way you'd vet any other third-party integration: check the publisher, check the reviews, and don't grant more access than the task requires.
Creator Economics: Publishing & Revenue Share
If you only used Agent.ai as a consumer, you can skip ahead, but if you built and published agents there, this section is the one that actually affects your income, not just your workflow. Agent.ai never operated a public revenue-share program for its creators; agents were free to build and publish, which was great for discoverability but meant there was no direct path from "my agent is popular" to "my agent pays me." Caywork's model is built around the opposite assumption.
Caywork's 75/25 Split
Every transaction on Caywork's marketplace pays the creator 75% of the sale, with the platform keeping a 25% fee (plus generating one platform credit per transaction). That's a real, structural payout, not a bonus program or a grant, but the default economics of publishing on the platform. If you had agents on Agent.ai that people actually used regularly, rebuilding them on Caywork is the first time that usage could translate into direct income rather than just reputation.
Why HubSpot Agent Hub Doesn't Offer This
HubSpot's Agent Hub, the product that effectively absorbed Agent.ai's technology, is built for HubSpot customers building agents inside their own CRM; it's not an open, third-party creator marketplace, and it was never positioned as one. That's a meaningful gap for anyone who thought of themselves as a builder first and a HubSpot user second: there's no equivalent "publish and get paid" path inside HubSpot's ecosystem, which is exactly the gap an open marketplace with a real revenue share is built to fill.
Enterprise Readiness: SSO, Multi-Workspace (V22)
For teams evaluating a platform rather than individuals, security and account structure often decide the shortlist before pricing even gets discussed. This is an area where the intel from the two ecosystems is genuinely asymmetric right now: most standalone agent platforms in this comparison set don't have a documented enterprise security story at all. Caywork's most recent release is a direct answer to that gap, even if it's still early.
Caywork's V22 release added SAML-based single sign-on (SSO), giving enterprise IT teams a standard way to manage authentication centrally, along with the ability to create and switch between isolated workspaces (separate chats, agents, and settings per workspace, with workspace-level theming). If your team runs multiple brands, clients, or departments through one account, that workspace isolation is the feature to test first.
How to Migrate from Agent.ai to Caywork
Knowing the two platforms are different is one thing; actually moving your workflows over is the part that costs real time, so it's worth setting expectations before you start. There's no import tool that reads an Agent.ai export and rebuilds it on Caywork automatically; that kind of cross-platform migration tooling essentially never exists between competitors, and Agent.ai's own shutdown notice confirmed there's no official migration path at all. What you can do is work through your old setup deliberately, agent by agent, using Caywork's marketplace and no-code builder to reconstruct the logic rather than just the outcome.
Start by listing every agent or workflow you actually used in the last 90 days, not everything you ever built, just what was load-bearing. Check the marketplace for a direct equivalent first (the migration map is built for exactly this lookup), and for anything without a close match, rebuild it with the visual builder using Loop, Condition, and Delay nodes to recreate any multi-step logic. Reconnect your integrations through the Connection Manager, and if you were monetizing agents on Agent.ai, register as a creator before you publish so the 75/25 split applies from your first sale.
Agent.ai vs. Caywork: Final Verdict---Who Should Switch
There's no single right answer here, so this is less a verdict than a way to sort yourself quickly. If you were a heavy Agent.ai user who mostly consumed agents for research, sales prospecting, or content drafting, Caywork's Marketing and Development categories cover a lot of that ground already, and the pay-once credit model means you're not committing to a subscription just to test the waters. If you were a creator who built and published agents, the calculus is even more direct: Caywork is one of the only remaining open marketplaces with a real, structural revenue share, which is something Agent.ai never had, and HubSpot's Agent Hub isn't built to offer. If your workflows were deeply HubSpot-native to begin with, Agent Hub may genuinely be the simpler landing spot, but if you valued Agent.ai precisely because it wasn't locked to one CRM, Caywork is the closer philosophical match.
FAQ
Is Caywork really free to start?
There's no monthly free tier, but you can browse the entire marketplace and see exactly what each agent does and costs before spending a credit. The Starter Pack is $5 for 100 credits, which is enough to test several agents before deciding whether to buy more.
How long does migration take?
It depends entirely on how many workflows you're rebuilding and how complex they were. A single-agent workflow can usually be recreated in under an hour; a multi-step Agent Teams setup will take longer since you're rebuilding the logic with Loop, Condition, and Delay nodes rather than importing it.
Can I import my Agent.ai data?
No, there is no official export or import tool between the two platforms, and Agent.ai's own closure notice confirmed no migration path was announced for creator listings. Plan to rebuild rather than transfer.
Next Steps
If you're ready to move, the fastest path is to browse the marketplace for your most-used Agent.ai workflows first, then check the complete migration map for anything you can't find. If you built and monetized agents on Agent.ai, read up on Caywork's creator economics before your first publish so the 75/25 split is working for you from day one.
References
- Caywork: Agent.ai Closes on August 22, 2026: https://caywork.com/learn/blog/agent-ai-closes-august-22-2026
- Caywork Pricing: https://caywork.com/pages/pricing
- Caywork: Caywork V22 Is Here: https://caywork.com/learn/blog/caywork-v22-is-here
- Caywork Agents Marketplace: https://caywork.com/agents
- Boston Globe: HubSpot cofounder Dharmesh Shah rolls out marketplace for AI 'agents' (Jan 31, 2025): https://www.bostonglobe.com/2025/01/31/business/hubspot-dharmesh-shah-ai-artificial-intelligence-agents/
- MarTech --- HubSpot July 2026 updates: Agent Hub arrives: https://martech.org/hubspot-july-2026-updates-agent-hub-arrives-and-more-ways-to-control-what-you-see/
- CMSWire: HubSpot Debuts Agent Hub to Unify AI Agents: https://www.cmswire.com/customer-experience/hubspot-debuts-agent-hub-to-unify-ai-agents/
- BuilderPack (agent.ai now redirects here, a HubSpot Labs product): https://builderpack.com/
